Reading bitset<0> from istream always a failure?

Question

I was recently working with std::bitset<0> and noticed something interesting with operator>>:

The "specification" states that:

If no characters are extracted, is.setstate(ios_base::failbit) is called.

Does this still apply when "all possible N characters" (with N == 0) have been read? If so, does this imply that trying to read from an istream into a bitset<0> must always set failbit?

The following unit test exhibits no consistent behavior across the compilers I had at hand:

BOOST_AUTO_TEST_CASE(io0) {
    try {
        std::stringstream ss{"1"};
        ss.exceptions(std::ios::badbit | std::ios::failbit);
        std::bitset<0> b;
        ss >> b;
        BOOST_TEST(false);
    } catch(...) {
        BOOST_TEST(true);
    }
}

MSVC 2019 sets the failbit, whilst GCC 7.3.0 does not!

Davis Herring · Accepted Answer · 2019-08-17 06:30:39Z

2

The specification (no scare quotes) said the same thing (when this question was asked), so yes. (Of course, skipping whitespace first may provoke some other kind of failure instead.) I thought this a clear error in the standard, so I filed an issue which has since been resolved for C++20.

edited Aug 17, 2019 at 6:30

answered Apr 4, 2019 at 1:36

Davis Herring

42.7k4 gold badges61 silver badges94 bronze badges

Sign up to request clarification or add additional context in comments.

MFH Over a year ago

My only observation from reading through the linked draft is that a similar case (std::array<T, 0>) is explicitly mentioned in the standard, whilst bitset<0> is not mentioned whatsoever...

Davis Herring Over a year ago

@MFH: So it sounds like 0 is valid (array has to mention it because of its otherwise-invalid exposition-only array member).

Collectives™ on Stack Overflow

Reading bitset<0> from istream always a failure?

1 Answer 1

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related